Field of the Invention
The present invention relates to an encoding apparatus, an encoding method, and a non-transitory computer-readable storage medium.
Description of the Related Art
H.264/MVC (Multi-view Video Coding) is known as a technique for encoding multi-viewpoint video data. According to H.264/MVC, it is possible to efficiently compress video data by not only referring to another frame captured by the same imaging unit (motion prediction) but also referring to a frame captured at another viewpoint (parallax prediction).
Video transmission using an IP (Internet Protocol) network such as the Internet has become widespread. RTP (A Transport Protocol for Real-Time Application, RFC 3550, IETF) as a protocol for transmitting encoded data of audio or a moving image in real time has been used for such video transmission. RTP assumes the use of UDP (User Datagram Protocol) with a high transfer rate in the transport layer. Although the transfer rate of UDP is high, UDP has no solution to a packet loss or does not ensure a transmission time. It is, therefore, necessary to deal with a communication error in a layer higher than UDP. If an error occurs in a video data packet encoded by not only referring to other frames but also referring to frames captured at other viewpoints, disturbance in the video due to the error propagates in the time direction at other viewpoints.
Japanese Patent Laid-Open No. 2010-506530 discloses, as a method of preventing error propagation, an intra-refresh technique of periodically inserting an intra-prediction encoded intra-frame, and a technique of alternately applying the intra-refresh technique to videos of a plurality of viewpoints.
In the technique described in Japanese Patent Laid-Open No. 2010-506530, however, if an error occurs in the first half of a cycle in which an intra-frame is inserted, it is impossible to recover from video disturbance due to error propagation for a long time.
There is provided a method of shortening the cycle to frequently insert an intra-frame in order to quickly recover from video disturbance due to error propagation. However, with this method, the data amount of the intra-frame is large, and frequent insertion of such intra-frame increases the transmission data amount. This, therefore, causes a further packet loss due to network congestion.
The present invention provides a multi-viewpoint video data encoding technique which allows early recovery from error propagation. Alternatively, the present invention provides a multi-viewpoint video data encoding technique which enables to reduce the transmission data amount.